sign in sign up
<<
Home , Actophilornis, Attagis, Charadriiformes, Cursorius, Egyptian plover, Ibisbill

College of Saint Benedict and Saint John's University DigitalCommons@CSB/SJU

Biology Faculty Publications Biology

2-1995

Phylogenetic reanalysis of Strauch's osteological data set for the

Philip C. Chu College of Saint Benedict/Saint John's University, [email protected]

Follow this and additional works at: https://digitalcommons.csbsju.edu/biology_pubs

Part of the Biology Commons, Commons, and the Commons

Recommended Citation Chu, PC. 1995. Phylogenetic reanalysis of Strauch's osteological data set for the Charadriiformes. The Condor 97(1): 174-196.

Published as Phylogenetic reanalysis of Strauch's osteological data set for the Charadriiformes. The Condor 97(1): 174-196. ©1995 by The Cooper Ornithological Society. Copying and permissions notice: Authorization to copy this content beyond fair use (as specified in Sections 107 and 108 of the U. S. Copyright Law) for internal or personal use, or the internal or personal use of specific clients, is granted on behalf of The Cooper Ornithological Society for libraries and other users, provided that they are registered with and pay the specified eef via Rightslink® on JSTOR (http://www.jstor.org/r/ucal) or directly with the Copyright Clearance Center, http://www.copyright.com. TheCondor97:174-196 0 The Cooper Ornithological Society 1995

PHYLOGENETIC REANALYSIS OF STRAUCH’S OSTEOLOGICAL DATA SET FOR THE CHARADRIIFORMES

PHILIP c. CHU Department of Biology and Museum of Zoology The University of Michigan, Ann Arbor, MI 48109

Abstract. Strauch’s (1978) compatibility analysisof relationshipsamong the shorebirds (Charadriifonnes) was the first study to examine the full range of charadriifonn taxa in a reproducibleway. SubsequentlyMickevich and Parenti (1980) leveled seriouscharges against Strauch’s characters,method of phylogenetic inference, and results. To account for these charges,Strauch ’s characterswere re-examined and recoded, and parsimony analyseswere performed on the revised matrix. A parsimony analysison 74 taxa from the revised matrix yielded 855 shortesttrees, each length = 286 and consistencyindex = 0.385. In each shortest tree there were two major lineages,a lineageof -likebirds and a lineageof - like ; the two formed a monophyletic group, with the (Alcidae) being that group’s sister taxon. The shortest trees were then compared with other estimates of shorebird re- lationships, comparison suggestingthat the chargesagainst Strauch’s results may have re- sulted from the Mickevich and Parenti decisions to exclude much of Strauch’s character evidence. Key words: Charadrilformes; phylogeny; compatibility analysis: parsimony analysis; tax- onomic congruence.

INTRODUCTION Strauch scored 227 charadriiform taxa for 70 The investigation of evolutionary relationships characters. Sixty-three of the characters were among shorebirds (Aves: Charadriiformes) has a taken from either the skull or postcranial skel- long history (reviewed in Sibley and Ahlquist eton; the remaining seven involved the respec- 1990). Almost all studies used morphology to tive origins of three neck muscles, as published make inferences about shared ancestry; infer- in Burton (1971, 1972, 1974) and Zusi (1962). ences were made using an intuitive method. These data were analyzed using the method of character compatibility (Estabrook 1972; Esta- Much of the literature of systematicsdeals with the brooket al. 1975,1976a, 1976b; McMorris 1975; identification of characterswhich are good estimators Es&brook et al. 1977). of phylogenetic history. Early systematistshad little In compatibility analysis, characters are treat- more than their own insights to help them choosethe characterswhich best indicate relationships. The sta- ed as partially-ordered sets; that is, each char- bility of much of zoological classificationis testimony acter is viewed as a set of states, with the states to their goodjudgment in their choices.Their methods, being ordered by some hypothesis of character however, have made it difficult or impossiblefor others evolution (Fig. 2). Strauch generated hypotheses to follow or repeat the stepsfrom observationsof spec- of character evolution using a common = prim- imens to the statements of relationship among taxa itive criterion. The most common state was de- (Strauch 1978:269). termined with reference to the Charadriiformes Appropriately, Strauch’s own (1978; Fig. 1) in- alone. For many characters, the most common vestigation of charadriiform relationships was state in two outgroups- the cranes and their rel- explicit in both its assumptions and method of atives () and the pigeons (Columbi- clustering taxa, and is therefore repeatable. In- formes)-was identified as well, but if the most deed, among morphological studies examining common outgroup state differed from the most the full range of charadriiform taxa, only that of common charadriiform state, then the latter was Strauch meets the criterion of repeatability; a coded as primitive (e.g., characters 38, 45, and second such study is currently under way 56). (McKitrick, unpubl. manuscript). Many of Strauch’s characters had more than two states. For each of these an ordered trans- formation series was constructed, though Strauch said (1978: 277) that his hypotheses oforder were ’ ’ Received 4 April 1994. Accepted 14 October 1994. sometimes only a guess.

11741 REANALYSIS OF CHARADRIIFORM PHYLOGENY 175

Cladorhynchus leucocephalus, Recurvirostra

Himan topus lbidorhyncha struthersii Haematopus (Vanellinae ) true (Charadriinae), feltohyas australis thick-knees (Burhinidae) Pluvianus aegyptius Dromas ardeola Pluvianellus socialis Chionis alba (Cursoriinae)

(Glareolinae) 1 ’ - I I (Sterninae) (Larinae) (Stercorariinae) Hydrophasianus chirurgus,

Metopidius indicus, Actophilomis, lrediparra gallinacea, Microparra capensis (Thinocoridae) painted (Rostratulidae) (Phalaropodinae) , Heteroscelus, Catoptrophorus semipalmatus Numenius, Bartramia longicauda Prosobonia cancellata Arenaria , Aphriza virgata, , Eurynorhynchus pygmaeus, Micropalama himantopus, Tryngites subruficollis, Philomachus pugnax, Limicola falck7ellus

Limnodromus Xenus cinereus

Limosa Coenocorypha aucklandica Lymnocryptes minimus Philohela minor, Scolopax auks (Alcidae) FIGURE 1. Strauch’s hypothesisof shorebird relationships. The tree is modified from his (1978) fig. 36; modificationswere made as specifiedin his fig. 2.

As its name suggests,the character-compati- senting character-statecombinations not in the bility method requires that the compatibility of study collection are discarded and the result is a charactersbe assessed.To determine whether or non-reticulate tree-any non-reticulate tree- not two charactersare compatible, their cartesian then the characters are compatible. Character product is calculated. The result is a new set that compatibility thus refersto the conditionin which can be visualized as a lattice; the lattice has a the sequenceof character-statetransformations vertex for every possible combination of states hypothesizedfor two or more characterscan be from both characters (Fig. 3). If vertices repre- accommodated by a single phylogenetictree. 176 PHILIP C. CHU

a a

FIGURE 2. Charactersas partially ordered sets. In the ball-and-stick diagrams shown here, each ball is a characterstate; the balls are connectedso as to depict hypothesesof character-statepolarity and order. In diagram A, state a is primitive and state b is derived. In diagram B, a is primitive; b is derived from a; and both c and d are derived independentlyfrom b.

The likelihood of finding a large set of mutually be determined, there is no underlying biological compatible charactersis taken to be small unless validity” (1980: 109). A secondcharge addressed compatibility results from shared evolutionary the particulars of Strauch’s character-state cod- history. Thus, large setsof compatible characters ing, ordering, and polarization; a third addressed provide better evidence of shared history than Strauch’s results,which were claimed to be “rad- do small sets,and permit greater confidence that ically different from all previous published hy- the tree capable of accommodating their respec- potheses” (1980: 109). tive transformation series is the true genealogy. Mickevich and Parenti, disagreeingwith some Initially, the largestset of mutually compatible of Strauch’s coding procedures,rejected 35 of his characters is determined. This initial analysis 70 characters, then analyzed the remaining 35 typically resolvesthe deepestpart of the tree, but using the computer program WAGNER 78 (Far- fails to resolve relationships at less general hi- ris 1978). Two shortest Wagner networks were erarchical levels. found; thesewere rooted with a hypothetical an- Within each group identified by the initial cestor constructed from characters for which analysis, the resolution of relationships is accom- Strauch described an outgroup state (Lundberg plished by determining which charactersvary in 1972). The consensustree calculated from these that group and then identifying from those vari- two shortest networks is shown in Figure 4. able charactersa largest clique. The processmay Unfortunately, the Mickevich and Parenti be repeated in smaller and smaller groups until reanalysis is itself not entirely satisfactory. As the tree is suitably resolved. indicated by the character descriptions below, Mickevich and Parenti (1980) leveled serious the Mickevich and Parenti rationale for rejecting charges against the Strauch study. One charge characters is heavy-handed, unevenly applied, addressed Strauch’s use of the compatibility and occasionally erroneous. For example, nine method. Mickevich and Parenti argued that characters(11, 14-17, 33, 36, 39, and 59) were compatibility analysis, by using only mutually rejected because taxa with multiple states were compatible characters to determine tree topol- given single state assignments.In most instances, ogy, selectively ignores characters that are in- however, the number of multistate taxa was compatible; they wrote, “We see no philosoph- small-only one out of 227 taxa for character ical distinction between the practice of 33 -and in one case(character 17) there were no disregarding data because of individual prefer- variable taxa. In another example, two characters ence and discarding data becauseof some formal (8 and 46) were rejected because some of their mathematical criterion for which, as far as can states were defined solely on the basis of con- REANALYSIS OF CHARADRIIFORM PHYLOGENY 177

b 1 3

D? 2 V

a

J bl b3

b2

al a3

Ma2

bl b3 bl b3 bl b3

b2 b2

69 @ al a3 al a3

* Ya2 a2 a2 C FIGURE 3. Character compatibility (modified from Strauch’s [1978] fig. 4). The Cartesianproduct of two characters(top row) is calculated,that product can be visualized as a lattice (middle row), with vertices in the lattice (e.g., al, a2, bl, and etc.) representingevery possible combination of characterstates. Combinations of statesnot observed in any of the taxa analyzed (shown in the bottom row as hatched circles) are then deleted from the lattice. If this procedure results in a non-reticulate tree, the charactersare compatible (3A and B); if it does not (3C), the charactersare incompatible. gruence with other character evidence; the same those skuas into two lineages; yet all four are was true for an additional five characters (4, 3 1, invariant for the 35 characters that Mickevich 33, 39, and 66), but Mickevich and Parenti did and Parenti analyzed. Four taxa cannot be not note it. grouped into two lineages when those taxa are Moreover, at least one example was found in scored as being identical. which the Mickevich and Parenti results are not In sum, although Mickevich and Parenti consistent with their revision of the Strauch pointed out flaws in the Strauch study, their cri- (1978) data matrix. The Strauch matrix contains, tique and subsequent reanalysis are flawed as among other things, four skuas (Stercorariinae). well. Consequently, in the present study I recode The Mickevich and Parenti consensus tree groups all of Strauch’s characters and analyze the revised 178 PHILIP C. CHU

Rynchops 1 Rynchops 2 gulls (Larinae) 1 gulls (Larinae) 2 I I terns (Sterninae) 1 terns (Sterninae) 2 skuas (Stercorariinae) 1 skuas (Stercorariinae) 2 I --I! mursars (Cursoriinae) pratincoles (Glareolinae) 3 pratincoles (Glareolinae) 2 pratincoles (Glareolinae) 1 Pluvianus aegyptius thick-knees (Burhinidae) -l---eDmmas ardsola Chionis alba Pluvianehus socialis plovers (Charadriinae), Pehohyas australis lapwings (Vanellinae) 1 lapwings (Vanellinae) 2 Haematopus fbicbrhyncha struthersii Cladorhynchus leucocephalus, Recurvirostra 1 Cladcrhynchus leucocephafus, Racurvirostra 2 Himanropus A&is, Aphriza virgata, Calidds, Eurynorhynchus pygmaeus, Micropalama himanmpus, Tryngites subruficollis,Philomachus pugnax, Limicola fafcinellus: 7

Aranaria Actitis, Aphriza vfrgata, Calidris, Eurynorhynchus pygmaeus, Micropalama himantopus, Tryngites subrufimllis, Philomachus pugnax, Limicola falcinellus: 2

Limnodromus Limosa Xenus cinareus Numenius, Dartramia longicauda phalaropes (Phalaropodinae) Tringa, Heteroscelus, Catoptrophorus semipalmatus 1 Prosobonia cancel/ala jacanas () painted snipe (Rostratulidae) Coenocorypha aucklandica Lymnocryptes minimus Philohela minor, Scolopax Gallinago lapwings (Vanellinae) 3 lapwings (Vanellinae) 4 auks (Alcidae) FIGURE 4. The Mickevich and Parenti estimateof shorebirdrelationships. The topologyshown is an Adams consensustree, calculatedfrom the two minimum-length resolutionsthat Mickevich and Parenti found.

matrix generated by the recoding process. My ing as many potentially informative data as they objective is to produce a hypothesis of charad- did. Such a hypothesis can then be used as a riiform phylogeny that accounts for the Mick- framework for examining character evolution evich and Parenti critique, but without discard- within the Charadriiformes (e.g., Chu 1994). REANALYSIS OF CHARADRIIFORM PHYLOGENY 119

REVISION OF STRAUCH’S DATA MATRIX ation referring to presenceor absenceof the an- terolateral condylar groove. However, Strauch CHARACTERS REJECTEDBY MICKEVICH AND did not assignthe primitive state to any glareolid PARENT1 taxon. Instead, he assignedall glareolids derived The 35 characters rejected by Mickevich and state C, a legitimate state assignment since state Parenti are discussed in turn below. For each C was defined not by the variable grooving, but characterthe Mickevich and Parenti rationale for by a unique, invariant orientation of the condyle rejection is examined and either supported,mod- itself. ified, or not supported. Coding:in Strauch,a singlecharacter with three states, A, B, and C. I recoded this character as Character 8: form of the bill tip two new characters.The first character, describ- Mickevich and Parenti argued that the derived ing the anterolateral groove of the medial con- states for character 8 were defined on the basis dyle, had two states corresponding to Strauch’s of presumedconvergence, i.e., that a singleshared states A and B (i.e., to presenceand absence of state was recoded as two or more states when the anterolateral groove, respectively). All taxa preliminary analysis suggestedthat the shared to which Strauch assignedstate C were question- statewas homoplasious.This is incorrect: Strauch marked, since Strauch doesnot make clear which did change the coding for this character during of those taxa are variable with respect to the the course of his investigation, but only because groove. The second character, describing orien- his initial coding scheme proved inadequate to tation of the medial condyle, had two states,one describe the diversity in shorebird bill tip mor- corresponding to Strauch’s statesA and B (low- phology, not becauseof any initial analysis sug- ered and untwisted), the other to his stateC (raised gesting homoplasy. and twisted). Coding: in Strauch, six states, A through F; unchanged here. Character 14: maxillopalatine strut A Mickevich and Parenti pointed out, correctly, Character 9: shapeof the cross-sectionof the dor- that variable taxa were assignedthe statethought sal bar of the upperjaw to be primitive. However, only 19 of the 140 Character 9 was rejected by Mickevich and taxa assignedthe primitive state were variable; Parenti because its derived states are autapo- the variable taxa were cucullatusand morphic. However, of the three derived states all members of the genera Haematopus, Ibido- identified by Strauch, only one (state B) is an rhyncha, Himantopus, Cladorhynchusand Re- autapomorphy. Rejecting the character seems curvirostra.Instead of rejecting the characterout- excessive,because it has two derived statesthat right, I elected to give each variable taxon a are sharedin addition to the one that is not shared. multiple state assignment,thereby indicating that Coding: in Strauch, four states, A through D, some individuals had one state and some, the unchanged here. Additionally, I assigned mul- other. tiple statesto one taxon, Xenus cinereus;Strauch Coding: in Strauch, two states, A and B; un- assignedonly one state to Xenus, despite his in- changed here. dication that the speciesis variable. Character II: medial condyle of quadrate Character 15: maxillopalatine strut B Strauch recognized three states for character Character 16: maxillopalatine strut C 11: A, in which the anterolateral face of the me- As Mickevich and Parenti again correctly dial condyle bears a distinct groove; B, in which pointed out, variable taxa were assignedthe hy- there is no groove; and C, in which the groove pothesizedprimitive state for both characters15 is reduced or absent, but the condyle itself is and 16. For each character the number of vari- raised and twisted. State A was hypothesized to able taxa is no higher than nine (out of 227 taxa be primitive. in Strauch’s analysis); thus I did not reject either. Mickevich and Parenti rejected character 11 All of the variable taxa are in the Tringa. becausevariable taxa were assignednot multiple Unfortunately, however, Strauch says only that states but A, the presumed primitive state. “some specimens” of Tringa are variable (1978: Strauch recognized only the (cours- 299), raising the question, “Some specimens of ers and pratincoles) as being variable, the vari- which Tringa?” Since Strauch provides no an- 180 PHILIP C. CHU swer, I was forced to question-mark all members taxa, and rejected characters 24-30 because of of that genus for both characters. it. I retained the characters, but assigned char- Coding: in Strauch, each character has two acter states only to those taxa scored by either states, A and B; neither is changed here. Burton or Zusi. All remaining taxa (Appendix 1) were scored as having missing data. Character 17: maxillopalatine strut D Coding: in Strauch, characters24-29 had two Character 17 was rejected by Mickevich and stateseach, A and B; character30 had three states, Parenti because variable taxa were assignedthe A through C. In no casewas the coding changed. presumed primitive state. However, Strauch did not indicate that any taxa were variable for this Character 31: number of caudal vertebrae,omit- character. ting the pygostyle Coding: in Strauch, two states, A and B; un- This character was rejected by Mickevich and changed here. Parenti because of unresolved positional ho- mology. Rejection of this and similar characters Character 18: supraoccipitalforamina identified by Strauch suggestsa philosophical The supraoccipital foramina are the fonticuli disagreement with the use of meristic characters occipitalis of Baumel (1979) a pair of openings in phylogenetic analyses. I see no a priori reason located on either side of the cerebellar promi- to reject such characters, so long as the set of nence. Mickevich and Parenti rejected the fo- things to be counted can be clearly delimited, ramina as a potential source of historical infor- and so long as the sets themselves meet certain mation on the grounds that their “positional requirements for structural and positional sim- homology is unresolved” (1980: 112). However, ilarity. neither Strauch’s character description nor my Coding: in Strauch, six states, A, B, C, D, E, own examination of charadriiform skulls has and G. There is no state F listed. Both states D suggestedthat the foramina differ sufficiently in and G code for six caudal vertebrae, with state position or configuration to warrant rejection of G representing six vertebrae in the jacanas (Ja- the character on positional grounds. canidae) and state D, six vertebrae in any other Coding: in Strauch, two states, A and B; un- shorebird. Strauch defended this coding scheme changed here. by arguing, “All Jacanidae have five vertebrae Character 21: prearticular processof the man- except Hydrophasianuschirurgus, which has six. dible The sixth caudal vertebrae in chirurgus is as- Mickevich and Parenti rejected this character sumed to be derived, along with the greatly elon- because its presumed derived states are auta- gated tail of the ,from the normal five of pomorphic. However, the character has five the Jacanidae” (1978:305). states, only one of which (state D) is limited to The recoding of one character state as two is a single taxon. Because each of the remaining inappropriate if the two differ only by the group- four statesis shared, rejecting the character out- ings in which they are observed. Thus I treated right eliminates a considerable amount of infor- the two six-vertebrae states,D and G, as a single mation potentially useful for grouping. state, reducing the number of states to five: A, Coding: in Strauch, five states, A through E; B, C, D/G, and E. unchanged here. Character 32: number of neck vertebrae,omitting Characters 24-30: head and neck muscle char- those with unfusedcervical ribs acters Character 35: number of sternal costalprocesses These charactersdescribe the origins of three Again, Mickevich and Parenti rejected meris- neck muscles,as published in Burton (197 1, 1972, tic characters because of unresolved positional 1974) and Zusi (1962). However, neither Burton homology nor Zusi examined many of the taxa included in Coding: in Strauch, three states for character Strauch’s study, so Strauch used Burton’s and 32 (A through C) and four for character 35 (A Zusi’s data to assigneach unexamined taxon the through D); unchanged here. state shown by a presumed close relative that had been examined. Character 33: cervical vertebral strut Mickevich and Parenti recognized the impro- Character 33 was rejected by Mickevich and priety ofassigning characterstates to unobserved Parenti becausea single species(out of 227!) was REANALYSIS OF CHARADRIIFORM PHYLOGENY 181

variable and was assignedthe character state hy- was coded as a uniquely derived state” (1978: pothesized to be primitive. To correct this, I as- 309). Becauseit is inappropriate to use character- signedthe one species,Pluvianus aegyptius, mul- state distribution among hypothesized groups as tiple states. a basis for recoding characters, I reduced the Coding: in Strauch, three states,A through C. number of statesto two, one (A) for the presence Both states B and C indicate the presence of a of the foramen and the other (B/C) for its ab- cervical vertebral strut, state B in the lapwings sence. (Vanellinae) and state C in the thick-knees (Bur- hinidae). Strauch justified this redundancy by Character 45: ectepicondylar prominence of the humerus stating, “I originally coded the presenceof a strut Mickevich and Parenti rejected this character as the same state for the lapwings and the thick- becausethey claimed the coding description con- knees, but after an initial analysis of the cladistic tains a typographical error; however, I found no relationships within the Charadrii indicated that erroneous state assignments. there is no closerelationship between thesegroups I recoded the strut as two different states” (1978: Coding: in Strauch, two states, A and B; un- changed here. 306). Mickevich and Parenti pointed out (though Character 46: extensorprocess of the metacarpus not with respect to the cervical vertebral strut) The inappropriate coding procedure noted by that a character cannot be recodedjust because Mickevich and Parenti, and discussed below, the analysis of other characterssuggests that re- caused them to reject character 46. I chose not coding is warranted. To correct for such inap- to reject the character, though I did change the propriate coding decisions,I listed only two states way it was coded. for character 33, one (state A) for the absenceof Coding: in Strauch, nine states, A through I. a cervical vertebral strut and the other (stateB/C) States B and I describe wing knobs, with state B for its presence. referring to wing knobs in Pluvianus aegyptius Character 36: medial sternal notch and state I, to knobs in Chionis alba and Pluvi- According to Strauch, some (Scol- anellus socialis. Similarly, statesC and F describe opacidae) were variable for this character. More- wing spursin the lapwings, with state F referring over, some were intermediate, with a medial fo- to wing spurs in cayanus and state C, ramen instead of a medial notch. Strauch’s to spurs in all other taxa. In defense of response was to code any variable or interme- these coding decisions, Strauch wrote, diate speciesas if it had the notch. I do not know The wing knobs in Pluvianellus. Chionis and Pluvianus which sandpiperswere variable or intermediate, were originally coded as the same state. Later it became and so was forced to question-mark all of them, clear that evidence from other characters indicates that as if they were missing data. the knob in Pluvianus has been independently derived, Coding: in Strauch, two states, A and B; un- and the character coding was changed accordingly. In changed here. addition, other evidence indicates that Hoploxypterus [= Vanellus] cayanus is not a member ofthe Vanellinae Character 39: coracoidal foramen and that its wing spur has been derived independently Mickevich and Parenti rejected character 39 of that found in the lapwings (1978:3 13). because Strauch made a single state assignment In response, Mickevich and Parenti wrote, to three variable taxa (Charadrius melanops, “The preliminary recoding of a character be- chalcopterus, and R. africanus). lieved to be different because other characters Rather than reject the character becauseof how support this differenceis unjustified” (1980: 111). three out of 227 taxa were scored,I retained the I agree, and so reduced the number of character character and restored the three variable taxa as states from nine to seven: A, B/I, C/F, D, E, G, having multiple states. and H. State B/I refers to the presence of wing Coding: in Strauch, three states,A through C. knobs; state C/F, to the presence of claw-like Both states B and C referred to the absenceof a wing spurs. coracoidal foramen, C in the Glareolidae and B in all other shorebirds; Strauch says, “After pre- Character 50: synsacral strut liminary analyses indicated that [, Cur- A synsacral strut is the transverse brace that sorius, and Rhinoptilus] form a monophyletic extends from the crista iliaca intermedia (Boas group, the absence of a foramen in [that group] 1933) at the caudal margin of the acetabular 182 PHILIP C. CHU

foramen, to one of the fused vertebrae associated Character 56: intrapophyseal foramina of syn- with the synsacrum. The strut comprisesthe en- sacrum larged costal processesof a synsacral vertebra Character 56 refersto the degreeof ossification (Baumel 1979). Three of the four statesfor char- between the transverseprocesses ofadjacent syn- acter 50 identify a different vertebra bearing the sacral vertebrae. Mickevich and Parenti rejected synsacralstrut, whereasthe fourth state indicates it, as they did the previous character, on the that the strut was absent altogether. grounds of unresolved positional homology. Mickevich and Parenti cited unresolved po- However, what may be important here is not the sitional homology in rejecting character 50, per- exact location of each interapophyseal foramen, haps because the character’s states refer to dif- but whether or not synsacral ossification is suf- ferent vertebrae. However, the issueofpositional ficiently extensive to reduce the number of fo- homology can be resolved by defining the char- ramina present. If degree of ossification is the acter in a more inclusive way (as, e.g., enlarge- meaningful variable, then the locations of par- ment of the costal processesin a particular block ticular foramina are unimportant, and objectives of three synsacral vertebrae). about positional equivalence are obviated. Coding: in Strauch, four states, A through D, Coding: in Strauch, two states, A and B; un- unchanged here. changed here.

Character 51: second synsacral strut Character 59: fusion of the ischium and the pubis Mickevich and Parenti rejected character 5 1 Mickevich and Parenti rejected character 59 becauseof unresolved positional homology. I re- because variable taxa were assigned the pre- jected it as well, but for a different reason:Strauch sumed plesiomorphous condition; I rejected it indicated that several taxa are variable with re- as well, but for a different reason. Strauch wrote, spectto the secondstrut, but never identified the “Both [of the two states described for character variable taxa. Becausevariable taxa are not iden- 591 can be found in specimensof some species,” tified I had no choice but to omit the character yet he does not specify which species.Thus the altogether. greater problem is not that variable taxa were assignedone state or the other but that variable Character 55: foramina on ventral surface of il- taxa were not identified at all. ium Character 55 describesthe presenceor absence Character M-70: tendinal canals in the hypo- of foramina on the lateral part of the caudal iliac tarsus crest (the anterior margin of Holman’s [ 196 l] All of these characterswere rejected by Mick- renal bar). Mickevich and Parenti rejected this evich and Parenti because of unresolved posi- character because of unresolved positional ho- tional homology. However, Strauch’s hypotheses mology, presumably a reference to the absence of homology are outlined reasonably clearly; he of more specific positional criteria for any one provides a diagram of the hypothetical charad- foramen. However, an examination of the two riiform hypotarsus, as well as several labelled taxa for which Strauch found foramina, Stercor- examples that permit the reader to see how the ark pvmarinus and S. parasiticus, suggestedthat hypothetical hypotarsuswas used to identify ten- the foramina are consequent from pneumatiza- dinal canals in actual specimens. tion of the crista iliaca caudalis. Thus character Coding: in Strauch, for characters65 and 67, 55 might be more precisely described in terms three stateseach, A through C, for characters68, of the presence or absence of pneumatization. 69, and 70, two states each, A and B. None of Under theseconditions, positional similarity need the above were changed here. only be demonstrated for the crista iliaca cau- Strauch assigned character 66, tendinal canal dalis, which may or may not be pneumatized; it no. 3, a total of four states (A through D). State does not need to be demonstrated for any indi- C described the condition in which no canal was vidual foramen. visible, and state A, the condition in which the Coding: in Strauch, two states, A and B. The canal was an open groove. States B and D de- same two states are used here as well, but they scribed the condition in which the canal was refer to the presence/absenceof pneumatization, completely enclosed by bone, with state D rep- not the presence/absenceof foramina. resenting an enclosedcanal in the auks (Alcidae) REANALYSIS OF CHARADRIIFORM PHYLOGENY 183 and state B, an enclosedcanal in any other taxon. Character 49: proximal phalanx, digit III I cannot justify dividing the enclosed-canal con- Character 49 has two states, describing per- dition into two statesjust because it occurs in forate and imperforate proximal phalanges for two or more different groups;therefore, I reduced digit III. Strauch indicated that some specimens the number of statesto three, A, B/D, and C. of the Phalaropodinae @halaropes) and Glar- eolidae exhibited the perforate state. However, CHARACTERS ACCEPTED BY MICKEVICH he assigned all of them the imperforate state, AND PARENT1 arguing that, “in all cases the perforation is a Mickevich and Parenti accepted 3.5 of Strauch’s hole in a thin sheet of bone and appears to rep- characters. Two of these had to be recoded; in resent incomplete ossification or a damaged each case, the rationale for recoding is discussed specimen” (1978:3 14). below. In addition, six contained erroneous state Calling a perforate phalanx imperforate is ac- assignments. Erroneous state assignments were ceptable if the apparent perforation results from identified from handwritten correctionson a copy damage, but not if it is thought to result from of Strauch (1978) housed in the University of incomplete ossification: most, if not all, perfo- Michigan Museum of Zoology Division Li- rations in the phalanx are probably the result of brary. All errors were corrected and the correc- ontogenetic truncation of the ossification pro- tions verified by comparing Strauch’s (1978) ta- cess.Consequently, I treated the phalaropes and ble I with his (1976) Appendix III. glareolids as if some specieshad both the per- Character 4: angle betweenjugal bar and lateral forate and imperforate states,and since Strauch nasal bar never identified which phalaropes and glareolids Strauch recognized three states for character were variable I was forced to code all of them as 4. State A described the condition in which the if they were missing data for this character. angle between the jugal and lateral nasal bars is In addition to requiring recoding, character 49 less than or equal to about 60”; states B and C contains an incorrect stateassignment: Gygis alba described the condition in which the jugal bar- should be assignedstate B rather than state A. lateral nasal bar angle is about 70” or more. Character 53: number of lumbar vertebralpar- Regarding state C, Strauch wrote, “like [state] apophyses (B), found in Thinocoridae [seedsnipe],assumed For character 53, Rostra&la benghalensis to be a result of the short, finch-like bill and should be assignedstate A, not state C. possibly of independent origin” (1978:29 1). However, assumptionsof independent origin are Character 54: condition of posterior end of the not sufficient grounds for erecting an additional renal depression character state. As a result, I combined states B benghalensis should be assigned and C into a single state, B/C, reducing the num- state C, not state A. ber of states for character 4 to two: A and B/C. Character 25: origin of M. complexus on verte- MATERIALS AND METHODS bra 4 Before revision, the Strauch data matrix con- Strauch assigned Vanellus tricolor state C, but tained 70 characters.During revision, I omitted character 25 has only two states,A and B. Con- two of Strauch’s original characters(5 1 and 59), sequently I question-marked tricolor, as if it was split a third (character 11) into two, and changed missing data. the coding for an additional six (4, 31, 33, 39, 46, and 66). The net result was a reduction in Character 43: shape of the deltoid crest of the the total number of characters,from 70 to 69. humerus Recoding also rendered identical the charac- Character 43 contains an erroneous state as- ter-state descriptions for some of the 227 taxa in signment: Calidris can&us should be assigned the original Strauch matrix. Taxa with identical state A rather than state B. state assignments were combined under single Character 48: distal metacarpal symphysisof car- taxon labels, reducing the number of taxa in the pometacarpus revised matrix to 185; I then added a hypothet- For character 48, Gygis alba should be as- ical ancestor to the matrix, bringing the number signed state A rather than state B. of taxa to 186. 184 PHILIP C. CHU

States were assignedto the ancestor based on ployed in the second analysis were unchanged Strauch’s information about two outgroups, the from those used in the first, with three excep- Gruiformes and Columbiformes. However, state tions. assignmentswere made for only those characters To get a starting tree for branch-swapping, I in which all gruifonn and columbiform taxa sur- (1) did not use stepwiseaddition; instead, I used veyed had the same character state; ten charac- Tree 1 of the 100 shortesttrees found during ters (4, 6, 9, 17, 33, 34, 42, 43, 45, and 46) met the initial analysis. this requirement. Using a predetermined starting tree made The revised matrix (186 taxa and 69 charac- (2) multiple replications irrelevant. The multi- ters) is not printed here. Interested readers can ple-replication procedureis designedto com- reproduce it from Strauch’s matrix following the pensate for addition-sequence limitations, proceduresdescribed in Appendix 2. and no such limitations exist when starting I analyzed the revised matrix cladistically us- trees are obtained by a means other than ing the computer program PAUP 3.0s (Phylo- stepwiseaddition. As a consequence,I used genetic Analysis Using Parsimony; Swofford one replication rather than many. 199 1). All charactersin the matrix were weighted I set the maximum number of shortesttrees equally; multistate characters were treated as (3) saved to 2,500. unordered, and each multiple-state assignment, as if it representeda polymorphism. In addition, After 2,500 trees had been saved I allowed outgroup rooting was specified, with the hypo- branch-swapping to proceed for 141 hr, during thetical ancestor being designated as the out- which time branch-swapping was completed on group. 892 of the trees. The second analysis was then The use of unordered charactersis a significant terminated. departure from Strauch’s (1978) procedure In the first two analysesI placed arbitrary ceil- (Mickevich and Parenti [ 19801 do not indicate ings on the number of shortesttrees that PAUP whether they used Strauch’s hypothesesof char- could find, and in both cases the ceilings were acter-state order or not). I employed unordered reached. These results suggestedthat the popu- characters following Hauser and Presch (199 l), lation of shortesttrees was very large, and raised who argue that hypotheses of order should be concerns that even the 2,500-tree sample ob- determined from a cladogram, much as one uses tained in the second analysis was not represen- a cladogram to identify instances of homoplasy. tative of the population as a whole. In the initial PAUP analysis, shortesttrees were One way to reduce the population of shortest soughtusing a heuristic algorithm that employed trees is to delete taxa. Accordingly, I ran a third, a random addition sequenceand tree bisection- reduced-matrix PAUP analysis from which 112 reconnection branch-swapping; the addition/ of the 186 taxa had been deleted. Taxa were branch-swapping procedure was repeated ten likely to be deleted if they participated in con- times to compensatefor the limitations that each sensuspolytomies; for example, in the case of a addition sequence imposed. I also used the completely unresolved polytomy with six ter- MULPARS option, which savesall of the equal- minals, four of the six might be removed. Re- ly-parsimonious trees that are found and inputs moval of such terminals was a reasonable strat- them one by one into the branch-swapping pro- egy becauseit deleted taxa from cladesthat could cedure. not be resolved in an unambiguous way by the Becauseof the large number of taxa d com- available evidence. However, deletion was se- paratively small number of characters, I sus- lective with respect to shorebird , in- pected that many equally short trees would be suring that all suprageneric taxa recognized by found. As a result, I arbitrarily limited the num- Peters(1934), Jehl(l975), and Sibley et al. (1988) ber of shortest trees retained to 100 per repli- were represented by at least one speciesin the cation. Thus after ten replications as many as analysis. 1,000 trees could have been saved. The actual Parameters used in the third PAUP analysis number saved was 100, a consequence of only were identical to those usedin the initial analysis, one replication finding trees of minimal length. except that I limited the number of shortesttrees To searchfor additional minimum-length trees retained to 1,000 per replication. After ten rep- I ran a second PAUP analysis. Parameters em- lications up to 10,000 trees could have been re- REANALYSIS OF CHARADRIIFORM PHYLOGENY 185 tained; however, 855 were found during repli- The two major lineages were clustered into a cation 1 and none thereafter, suggestingthat all monophyletic group, and the auks were hypoth- shortest trees had been found. esized to be that group’s sister taxon. Within the lineage of sandpiper-like birds, the RESULTS jacanas, painted snipe, and sandpipers formed a The second of the first two analyses yielded a , with their sister taxon being the seedsnipe sample of 2,500 shortesttrees. Each required 40 1 (Fig. 6). However, the jacanas and painted snipe steps and had a consistency index (Kluge and were found to be derived sandpipers, not prox- Farris 1969) of 0.307. A strict consensus(Nelson imal outgroups to a monophyletic sandpiper 1979) of the 2,500 trees is shown in Figure 5 group; thus the sandpipers as usually conceived becauseof its previous use as a phylogenetic hy- (e.g., the Scolopacidae of both Peters 1934 and pothesis for the Charadriiformes (Chu 1994). Sibley et al. 1988) are hypothesized to be para- The third analysis (i.e., the reduced-matrix phyletic. analysis)yielded 855 shortesttrees, each 286 steps Relationships within the lineage of plover-like long and with a consistencyindex of 0.385; from birds were less clear. Strauch’s data supported these, strict (Fig. 6) and 50% majority-rule con- the existence of five plover-like , but did sensustrees (Margush and McMorris 198 1; Fig. not resolve the relationships between them (Fig. 7) were calculated. Given the large number of 6). These clades are the larids and the Crab Plo- taxa analyzed (74), the low consistencyindex cal- ver (Dromas ardeola); the coursers and pratin- culated for each shortest tree is expected; for a coles; the thick-knees and (Plu- 60-taxon analysis, Sanderson and Donoghue vianusaegyptius); the and Magellanic (1989) predict a consistency index of 0.349. Plover (Pluvianellus socialis); and the , av- The topological differencesbetween the results ocets, (Zbidorhynchastruthersii), oyster- of a full-matrix analysis (Fig. 5) and a reduced- catchers,lapwings, and true plovers. matrix analysis (Fig. 6) highlight the difficulties Resolution of relationships was particularly associated with data that are incapable of re- poor within the last of these clades. The lack of solving relationships among the taxa of interest. resolution was a direct consequence of my de- If existing data cannot resolve relationships, a cision to include the Ibisbill in the reduced-ma- very large number of equally simple resolutions trix analysis; deleting it reduced the number of are possible, only some of which will be found shortesttrees from 855 to 60 (Fig. 8). and subjected to branch-swapping; others will Whether the Ibisbill was included or excluded, neither be found nor swappedon. Conversely, if the majority of shortesttrees did not support the the existing data are capable of resolving rela- existenceof a monophyletic group comprised by tionships, a small number of shortest trees are the lapwings and true plovers (i.e., the Charad- possible, and the chancesare reasonable that all riidae sensu Peters 1934 or American Omithol- will be found, even with a heuristic search. For ogists’ Union 1983). Additionally, no shortest this reason I will limit further discussion to the tree supported the existence of a monophyletic reduced-matrix analysis. lapwing group: in the reduced-matrix analysis The reduced-matrix analysis (Fig. 6) indicated the closestrelative of Vanelluscayanus was Phe- two major clades of charadriiform taxa: gornis mitchellii, not Vanellus vanellusor V. chi- lensis. Strauch’s (1978) compatibility analysis (1) a lineage of sandpiper-like birds, including yielded a similar result in that it did not cluster the sandpipers, painted snipe (Rostratuli- cayanus with the other lapwings. dae), jacanas, and seedsnipe;and (2) a lineage of plover-like birds, including the DISCUSSION true plovers (Charadriinae), lapwings, oys- tercatchers (Haematopodidae), the Ibisbill The present analysis did not support sandpiper (Zbidorhynchastruthersii), stilts and , but it did recognize five sandpiper (), sheathbills (Chionidi- lineages (Fig. 6): the , tringine sandpipers, dae), the (Pluvianellusso- calidrine sandpipers, phalaropes, and . cialis), coursersand pratincoles, thick-knees, These lineages are similar to the scolopacidsub- the Crab Plover (Dromas ardeofa), and gulls families recognized in Lowe’s (193 1) seminal in- and their allies (). vestigation of shorebird morphology, but differ 186 PHILIP C. CHU

4 ’

h4med”S arqusr* I I ’ Nummdw ohaeoous

“*ndws ammws Vanelh~s malabaricus “a”Mlls rugutris Vanelb~s melatvpferus

FIGURE 5. Strict consensusof a sample of 2,500 shortesttrees found during analysisof the revised Strauch matrix. REANALYSIS OF CHARADRIIFORM PHYLOGENY 187

-

FIGURE 5. Continued. in two important respects. First, dowitchers and are viewed as closely re- (Limnodromus) emerge with the calidrine sand- lated calidrine sandpipers, was originally pro- pipers instead of the snipes, and godwits (Li- posed by Timmermann (1957a, 1957b) as a re- mosa) emerge with the dowitchers instead of the sult of his comparative examination of shorebird curlews.This hypothesisof relationships,in which ectoparasites. And second, a sister-group rela- 188 PHILIP C. CHU

Sco/opax rustic& Scolopax mira Tringa glareok Heteroscelus brevipes Heteroscelus incanus Tringa ery?hropus Tringa melanoleuca Tringa f/wipes Aphka Calidris tenuirostris Calidris alpha

Arenaria mef&ocepha/a Limnodromus griseus Limosa A Limosa fedoa Phalaropus tricolor Phalaropus A Bartramia Numenius A Numenius phaeopus Numenius arquata Numenius americanus rumicivorus AttaQis Stercorarius pomarinus Stercorarius parasiticus Larinae A Gelochelido!, ;;;;Fsps nlgra Glareola pratincofa Rhinoptilus cinctus A Cursorius cursor Cursorius rufus oedicnemus Burhinus senegalensis Burhinus magnirostris Burhinus capensis magnirostris Burhinus vermiculatus Pluvianus Pluvianellus Chionis alba Peltohyas Phegornis Vane//us cayanus Charadrius melanops Charadrius alexandrinus tenuirostris Charadrius dubius Vanellus~vanellus Vane//us chilensis Haematopus ostralegus Haematopus palliatus Haematopus fuliginosus Haematopus leucopodus lbidorh ncha Himan Yopus Cladorhynchus Recurvirostra novaehoflandiae Recurvirostra avosetta Ma aafge gryfle Hypothetical ancestor FIGURE 6. Strictconsensus of the 855 shortesttrees found duringa reduced-matrixanalysis of the revised Strauchmatrix. Eachshortest tree had a lengthof 286 stepsand a consistencyindex of 0.385. tionship between the (Armaria) and The enigmatic Crab Plover clustered with the (Aphriza virgata) was not supported, gulls and -like birds. This grouping was ini- though both clustered with the calidrine sand- tially hypothesizedby Yudin (1965), thoughLowe pipers. The relationships of Aphriza as hypoth- (19 16) had noted similarities between the Crab esized here are similar to those hypothesized by Plover and gulls half a century earlier. Jehl(l968). Another enigmatic taxon, the Egyptian Plover, REANALYSIS OF CHARADRIIFORM PHYLOGENY 189

Hydrophasianus Jacana iacana Rosfreda Nvcticrwhes Cben&rypha

L&~l$t%nagq Ga//i”ago paraguaiae Philohela Scolopax rusticola Sco/opax mira Tringa glareola Heteroscelus brevipes Heteroscelus incanus Tringa erythropus Tringa melanoleuca Tringa //a vipes Aphriza Calidris tenuirostris Calidris alpina Arenaria interpres Arenaria melanocephala Limnodromus griseus Limosa A Limosa fedoa Phalaropus tricolor PhdWODUS A Bartraniia Numenius A Numenius phaeopus Numenius arquata Numenius americanus Thinocorus rumicivorus Anaois Stercorarius pomarinus Stercorarius parasiticus Larinae A 100 /$w&ps nigra

Glareola pratincola Rhinoptilus cinctus CursoriusA Cursorius cursor Cursorius rufus Burhinus oedicnemus Burhinus senegalensis Burhinus magnirostris Burhinus capensis Esacus magnirostris Burhinus vermiculatus 1. Pluvianus Pluvianellus Chionis alba Peltohyas Phegornls Vane//us cayanus Charadrius melanops Charadrius alexandrinus tenuirostris Charadrius d&us Vane//us vane//us Vane//us chilensis lbidorhyncha Himantopus Cladorhvnchus Recurviiostra novaehollandiae Recurvirostra avosetta HaematoDus OSfra/egUS Haemafipus palliates Haematopus fuliginosus Haematopus leucopodus 100 - Uris aa/ge L Cepphus.grylle HypothetIcal ancestor

FIGURE 7. Fifty percent majority-rule consensusof the 855 shortest trees found during a reduced-matrix analysisof the revised Strauchmatrix. Numbers adjacentto each node are the percentageof shortesttrees having the resolution shown.

was grouped with the thick-knees. Strauch (1978) is “sharply differentiated from other Charadri- was the first to propose this relationship, but both iformes by precisely the same structural features his results and mine were presaged by Yudin’s as Burhinus.” (1965:224) observation that the Egyptian Plover Finally, two additional taxa of puzzling affin- 190 PHILIP C. CHU

I Peltohyas

Haematopus teucopvdus

Recurvirostra novaehollandiae Recurvirostra avosetta Charadr!us a/ex?ndrinus tenuiros:tris t ::::duLn”s

FIGURE 8. Strict consensusof the 60 shortest trees found when the Ibisbill (Ibidorhyncha struthersii) was deleted from the reduced-matrix analysis. Each shortesttree was of length 278 and had a consistencyindex of 0.392. REANALYSIS OF CHARADRIIFORM PHYLOGENY 191 ities, the sheathbills and the Magellanic Plover, catchers,stilts, avocets,and the Ibisbill were were grouped with one another, as suggestedby found to comprise a natural group. On the Jehl (1975) and later by Strauch (1978). Mickevich and Parenti topologies such a group would be an artificial assemblage(Fig. 4). COMPARISON OF TREES GENERATED IN THE STRAUCH, MICKEVICH AND PARENTI, AND PRESENT STUDIES COMPARISON WITH THE SIBLEY AND AHLQUIST ESTIMATE OF SHOREBIRD Strauch (1978; Fig. l), Mickevich and Parenti RELATIONSHIPS (1980; Fig. 4) and the present analysis placed Sibley and Ahlquist (1990) used the method of the auks at the node in the charadriiform DNA-DNA hybridization (Schildkraut et al. tree. The latter two found auks to be the sister 196 1; Britten and Kohne 1966; Shieldsand Straus taxon of all other charadriiform birds; the former 1975; Sibley and Ahlquist 198 1) to estimate re- was unable to resolve the position of the lationships of the birds of the world, including lineage, and so depicted it as part of a basal tri- 69 speciesof shorebirds.In general, they sampled chotomy. the different shorebird lineages less completely More-inclusive groupings among the non-auk than did Strauch (1978). They did, however, in- shorebirds were similar in the present study and clude both a greatervariety of auks and the Plains- that of Strauch, less so in that of Mickevich and wanderer (Pedionomus torquatus); the latter’s Parenti. Both Strauch and the present study clus- similarities to charadriiform taxa were noted only tered the non-auk shorebirds into two groups, recently (Olson and Steadman 198 1). one of sandpiper-like birds and the other of plo- The utility of comparisons with the Sibley and ver-like birds (Figs. 1,6). In contrast, Mickevich Ahlquist estimate may be questioned, because and Parenti found a lineage of lapwings to be the both Sibley and Ahlquist’s methods and their sister taxon to all other non-auk shorebirds (Fig. results have received much criticism (e.g., Brow- 4), with there being less resolution among the nell 1983; Cracraft 1987; Houde 1987; Sheldon latter than in either the Strauch analysis or my 1987; Sarich et al. 1989; Springer and Krajewski reanalysis(monophyly of the sandpiper-like birds 1989; Lanyon 1992; Mindell 1992). However, was supported unambiguously, but the position the Sibley and Ahlquist estimate is the only other of several lineages of plover-like birds was not comprehensive hypothesisof shorebird relation- resolved). ships. As such, it is an estimate with which any Additional comparisons between trees gener- charadriiform phylogeny should be compared. ated in the present study, the Strauch tree, and Sibley and Ahlquist’s evidence indicated that the Mickevich and Parenti trees are provided the Charadriiformes are made up of two groups, below. one of sandpiper-like birds and the other of plo- (1) My reanalysisof Strauch’s data indicated that ver-like birds (Fig. 9). Allowing for differences sandpipersare paraphyletic, with the jacanas in the taxa examined, membership in each group and painted snipe being derived members of was the same as in both Strauch’s study and the the group of sandpiper-like birds (Fig. 6). present analysis, with one exception: Sibley and The Mickevich and Parenti analysis yielded Ahlquist placed the auks within the group of plo- similar result (Fig. 4). Conversely, Strauch ver-like birds, with the auks being most similar found jacanas and seedsnipeto be the prox- to larids. imal outgroups for a monophyletic sandpip- In the Sibley and Ahlquist study the plover- er lineage (Fig. 1). like birds were themselves divided into two (2) Shortesttrees found in the presentstudy failed groups. One comprised the plovers, lapwings, to provide unambiguous support for the ex- stilts and avocets,, thick-knees, and istence of a monophyletic group made up of sheathbills.The other comprisedthe coursersand the lapwings and true plovers. Both Strauch pratincoles, the Crab Plover, auks, and larids. (Fig. 1) and Mickevich and Parenti (Fig. 4) The latter is identical in membership to a clade arguedagainst the monophyly of sucha group. supported by the Mickevich and Parenti analy- (3) In both this study (Fig. 6) and that of Strauch sis, except that Mickevich and Parenti placed (Fig. 1) the lapwings, true plovers, oyster- thick-knees in that clade as well (Fig. 4). 192 PHILIP C. CHU

fredipa~a g?//inacea ,44Cg/pffllS

fiosfrafula benohalensis Pedionomus fojquafus Affagis Thinocorus orbignyianus Thinocorus rumrcivorus scolopax GaNinago undulafa GaNinago megala Gallinago sfenura Numeflius Limosa Limnodromus Phalaropus Tringa flavipes Tringa melanoleuca Arenaria Micropalama himanfopus Tryn ifes subruficoffis Cahr? ris alba Calidris melanofos Calidris pusi//a Calidris minutilla Vane//us senegaffus Vane//us coronafus Vane//us leucurus

~zk~2izens

Vane//us cayanus Charadrius Anarhynchus honfalis Eudromias morinellus Oreopholus ruficollis Haemafopus unicolor Haemafopus osfralagus Haemafopus bachmani Haemafopus leucopodus Cladmynchus leumcephalus Himanfopus Recurvirostra Burhinus vermiculafus Burhinus capensis Burhinus bisfriafus Burhinus superciliads Chionis alba Glareola maldivarum S fikia isabella

Frafercula Lunda cirrhafa Cafharacfa Sfercorarius Rynchops niger ChEdonias nrger Sfema bargii Sfema hirundo Sfema hirundinacea marinus Larus laucoides Larus Ryperboreus Larus argenfafus

FIGURE 9. The Sibley and Ahlquist estimate of shorebird relationships. Topology was determined by UPGMA grouping of AT,,H values. See Sibley and Ahlquist (1990) for discussion regarding their choice of a distance metric and clustering method.

Sibley and Ahlquist found that the lapwings Strauch, Mickevich and Parenti, or the present (including Vunellus cayanus) are more similar to study. one another than any is to other charadriiform Finally, relationships within the Sibley and birds. In addition, they found the true plovers Ahlquist group of sandpiper-like birds were most to be the shorebirds most similar to lapwings. like those hypothesized by Strauch. The sand- Neither of these results were supported by pipers themselves emerged in a single cluster, REANALYSIS OF CHARADRIIFORM PHYLOGENY 193 whereas the jacanas, painted snipe, and seed- traduced,also involving fifty characters,and A(BC) snipe emerged outside of that cluster. is the most parsimonious hypothesisfor this sec- ond set.The datasets disagree. If parsimonyis the correctway to discernevidential meaning,this CONCLUSIONS resultdoes not in the slightestimpugn that meth- A single set of data was subjectedto phylogenetic od’scredentials. When different pieces of datapoint analysis three times: once by Strauch (1978) who in differentdirections, it is essentialto consider collected the data; once by Mickevich and Par- whatall the datasay. A natural“ principle of total enti, as part of their review of Strauch’s manu- evidence”enjoins us to find the most parsimo- script; and once here. The present study has sev- nioustree, relativeto all one hundredcharacters. eral advantages over its predecessors.First, it is Sober addressesthe situation in which different a parsimony analysis (like that of Mickevich and suitesof charactersyield different topologies,and Parenti) rather than a compatibility analysis (like concludesthat incongruence among topologiesis that of Strauch); the compatibility method has a commentary not on the method ofphylogenetic been strongly criticized for its exclusion of in- inference but on the character evidence em- compatible characters (Farris and Kluge 1979; ployed. However, his argument can be extended Mickevich and Parenti 1980; Sober 1988). Sec- to a second situation: caution must be exercised ond, the present study re-examines (and, where when assailing a method for yielding topologies appropriate, modifies) Strauch’s coding deci- different from those generated with other meth- sions in light of the Mickevich and Parenti re- ods, becausethe topologicaldifferences may stem view; at the same time, however, it discardsfew- from nothing more than an investigator’s choice er potentially informative data than Mickevich of characters. and Parenti did. And third, it provides a number If taxonomic congruence lends support to hy- of hypothesesabout relationships at the species potheses,then groupingspresent in several of the level that cannot easily be recovered from the hypothesesexamined here are better supported Mickevich and Parenti paper. Such information than groupingspresent in only one of them. Two is not available in Strauch’s (1978) paper, either, groups in particular are better supported in this but can be found in an earlier work of his (Strauch manner: the sandpiper-like birds (a group com- 1976). prised by the sandpipers,jacanas, painted snipe, Mickevich and Parenti (1980: 108) assert that and seedsnipe)and the plover-like birds (a group “[Strauch’s] interpretation of charadriiform phy- comprised by the lapwings, true plovers, oyster- logeny is radically different from all previous catchers, stilts and avocets, the Ibisbill, thick- published hypotheses, and may indicate pecu- knees, coursersand pratincoles, the Crab Plover, liarities of the [compatibility] method, rather than gulls and their allies, sheathbills, and the Mag- a close approximation of actual interrelation- ellanic Plover). The sandpiper-like group is hy- ships.” In so doing, they assertthat similarity of pothesized in Sibley and Ahlquist’s analysis of groupings (taxonomic congruence: Mickevich DNA-DNA hybridization data, in Strauch’s 1978) can be used to argue the relative merits of analysis of morphological characters,and in both compatibility and parsimony methods, much as my reanalysis of Strauch’s data and the Mick- it was once used to argue the relative merits of evich and Parenti reanalysis. The plover-like phenetics and (e.g., Sneath and Sokal group is hypothesized in all analysesexcept that 1973; Mickevich 1978). of Mickevich and Parenti, and might have been Sober (1988: 142) arguespersuasively that tax- hypothesized there as well, had Mickevich and onomic congruence “bears on the standing of Parenti not elected to exclude so much evidence hypotheses,not on the methodsused to select from their study. them.” He continues (1988: 142-143): Of course, as a source of support for hypoth- esizedgroupings, congruence with additional hy- pothesesis a poor substitute for congruencewith Let us suppose,just for the sakeof argument,that additional characters. It is the addition of new cladistic parsimony is the best method for assess- ing the evidential meaning of character distribu- characters, and not a comparison of trees, that tions. Let an initial set of fifty charactersbe such will prove the final arbiter in any discussionover that (AB)C is the best hypothesis,when parsimony which estimatesof relationships are most strong- is used to analyze the data. A new data set is in- ly supported. 194 PHILIP C. CHU

ACKNOWLEDGMENTS the Blackiths’ data on orthopteroid . Syst. Zool. 261269-276. W. L. Fink and M. C. McKitrick kindly lent their ex- FARRIS,J. S. 1978. WAGNER 78 program package. pertise in phylogeneticanalysis, and M. C. McKitrick, State Univ. of New York at Stony Brook, NY. D. P. Mindell, R. B. Payne, R. W. Storer, J. G. Strauch, FARRIS,J. S., AND A. G. KLUGE. 1979. A botanical Jr., and two anonymousreviewers provided numerous clique. Syst. Zool. 28:4Oti 11. helpful comments on previous versions of this manu- HAUSER,D. L., AND W. P~EXH. 1991. The effect of script. This study is submitted in partial fulfillment of ordered characters on phylogenetic reconstruc- the requirementsfor a Doctor of Philosophy degreeat tion. Cladistics 7~243-265. the University of Michigan. HOLMAN, J. A. 1961. Osteology of living and New World (Aves, ). Bull. Flor- LITERATURE CITED ida State Mus. 6:131-233. AMERICANORNITHOLOGISTS UNION.’ 1983. Check- HOUDE, P. 1987. Critical evaluation of DNA hy- list of North American birds, 6th ed. Allen Press, bridization studiesin avian systematics.Auk 104: Lawrence, KS. 17-32. BAUMEL,J. J. 1979. Osteologia, p. 53-122. In J. J. JEHL,J. R., JR. 1968. The systematicposition of the Baumel, A. S. King, A. M. Lucas, J. E. Breazile, Surlbird, Aphriza virgata. Condor 70:206-2 10. and H. E. Evans [eds.],Nomina anatomicaavium: JEHL,J. R., JR. 1975. Pluviane1Iu.s socialis: biology, an annotated anatomical dictionary of birds. Ac- ecology, and relationships of an enigmatic Pata- ademic Press,New York. gonian shorebird. Trans. S. Diego Sot. Nat. Hist. BOAS,J.E.V. 1933. Kreuzbein, Becken, und Plexus 18~25-74. lumbrosacralis der Vogel. D. Kgl. Danske Vi- KLUGE, A. G., ANDJ. S. FARRIS. 1969. Quantitative densk. Selsk. Skrifter, naturvidensk. og Mathem. phyleticsand the evolution of anurans.Syst. Zool. Afd. 9th ser. 5:1-59. 18:1-32. BRITTEN,R. J., AND D. E. KOHNE. 1966. Nucleotide LANYON,S. M. 1992. Review of, “Phylogeny and sequencerepetition in DNA. CarnegieInst. Wash. classificationof birds. A study in molecular evo- Yearbook 65:78-106. lution.” Condor 94:304-307. BROWNELL,E. 1983. DNA/DNA hybridization stud- LOWE,P. R. 1916. Studies on the Charadriiformes. ies of muroid rodents: symmetry and rates of mo- IV. An additional note on the sheath-bills: some lecular evolution. Evolution 37: 1043-l 05 1. points in the osteologyof the skull of an embryo BURTON,P.J.K. 1971. Comparative of head Chionarchus “minor” from Kerguelen. V. Some and neck in the Spoon-billed Sandpiper, Eury- notes on the Crab-Plover (Dramas ardeola Pay- norhynchuspygmaeusand itsallies. J. Zool., Lond. kull). 10th ser. 4:3 13-337. 163:145-163. LOWE,P. R. 1931. An anatomical review of the “wad- BURTON,P.J.K. 1972. Some anatomical noteson the ers” (Telmatomorphae), with special reference to . Notomis 19:26-32. the families, sub-families, and genera within the BURTON,P.J.K. 1974. Feeding and the feeding ap- suborders Limicolae, Grui-Limicolae and - paratus in : a study of anatomy and ad- Limicolae. Ibis 13th ser. 1:712-77 1. aptations in the Charadrii. Trusteesof the British LUNDBERG,J. G. 1972. Wagner networks and ances- Museum (Natural History), London. tors. Syst. Zool. 21:398413. CHU, P. C. 1994. Historical examination of delayed MARGUSH,T., ANDF. R. MCMORRIS. 1981. Consen- maturation in the shorebirds (Aves: sus n-trees. Bull. Math. Biol. 43:239-244. Charadriiformes). Evolution 48:327-350. McMoaats, F. R. 1975. Compatibility criteria for CRACRAFT,J. 1987. DNA hybridization and avian cladistic and qualitative taxonomic characters,p. phylogenetics,p. 47-96. In M. K. Hecht, B. Wall- 399-415. In G. F. Estabrook red.], Proceedingsof ante, and G. T. Prance [eds.], Evolutionary biol- the Eighth International Conferenceon Numerical ogy.__ Vol. 21. Plenum Press. New York. Taxonomv. W. H. Freeman, San Francisco. CA. ESTABROOK,G. F. 1972. Cladistic methodology: a MICKEVICH,M: F. 1978. Taxonomic congruenceSyst. discussionof the theoretical basisof the induction Zool. 27:143-158. of evolutionary history. Ann. Rev. Ecol. Syst. MICKEVICH,M. F., ANDL. R. PARENTI. 1980. Review 3:427-l56. of, “The phylogenyofthe Charadriiformes (Aves): ESTABROOK,G. F., C. S. JOHNSON,JR., AND F. R. a new estimate usingthe method ofcharacter com- MCMORRIS. 1975. An idealized concept of the patibility analysis.” Syst. Zool. 29: 108-l 13. true cladisticcharacter. Math. Biosci. 23:263-272. M~NDELL,D. P. 1992. DNA-DNA hybridization and ESTABROOK,G. F., C. S. JOHNSON, JR., AND F. R. avian phylogeny:review of, “Phylogeny and clas- MCMORR~S. 1976a. A mathematical foundation sificationofbirds: a studyin molecularevolution.” for the analysis of cladistic charactercompatibil- Syst. Biol. 41: 126-134. ity. Math. Biosci. 29:181-187. NELSON,G. 1979. Cladistic analysis and synthesis: ESTABROOK,G. F., C. S. JOHNSON,JR., AND F. R. principles and definitions, with a historical note McMoants. 1976b. An algebraicanalysis of cla- on Adanson’s Familles des Plantes (1763-l 764). distic characters.Discrete Math. 16:141-147. Syst. Zool. 28: l-2 1. ESTABROOK,G. F., J. G. STRAUCH,JR., ANDK. L. FIALA. OLSON,S. L., AND D. W. STEADMAN. 1981. The re- 1977. An application of compatibility analysisto lationships of the Pedionomidae (Aves: Charad- REANALYSIS OF CHARADRIIFORM PHYLOGENY 195

riiformes). Smithsonian Contrib. Zool. No. 337: YUDIN, K. A. 1965. [Phylogeny and classification of l-25. the Charadriiformes.] Fauna USSR No. 9 1, Char- PETERS,J. L. 1934. Check- of the world, adriiformes 2(l), part l:l-256 [in Russian, trans- vol. II. Harvard University Press, Cambridge, MA. lated by L. Kelso]. SANDERSON, M. J., AND M. J. D~NOGHUE. 1989. Pat- ZUSI, R. L. 1962. Structural adaptations of the head terns of variation in levels of homoplasy. Evolu- and neck in the Rynchopsnigru tion 43:1781-1795. Lunnaeus. Publ. Nuttall Om. Cl. 3:1-10 1. SARICH, V. M., C. W. SCHMID, AND J. MARKS. 1989. DNA hybridization as a guide to phylogenies: a critical analysis. Cladistics 5:3-32. APPENDIX 1: SCHILDKRALJT, C., J. MA~MUR, AND P. Don. 196 1. TAXA NOT EXAMINED BY The formation of hybrid DNA molecules and their BURTON OR ZUSI use in studies of DNA homologies. J. Mol. Biol. 5:595-617. The following taxa were not examined by either Burton (1971, 1972, 1974) or Zusi (1962) for features of the SHELDON, F. H. 1987. Phylogeny of herons estimated from DNA-DNA hybridization data. Auk 104: head and neck musculature: Metopidius indicus;Mi- 97-108. croparra capensis; albinucha; Jacana ja- SHIELDS,G. F., ANDN. A. STRAUS. 1975. DNA-DNA cana; gayi; Nycticryphessemicollaris; all Nu- hybridization studies of birds. Evolution 29: 159- menius except phaeopusand arquata; Calidris mauri 166. and ptilocnemis; Arenaria melanocephala; Limnod- SIBLEY, C. G., AND J. E. AHLQUIST. 1981. The phy- romusscolopaceus;Limosafedoa; Coenocoryphaauck- logeny and relationships of ratite birds as indicated landica; all Gallinago except stenura, gallinago, and by DNA-DNA hybridization, p. 301-335. In paraguiae; Scolopax mira; Catharacta ; all Ster- all gulls G.G.E. Scudder and J. L. Reveal [eds.], Evolution corarius; (Gabianusscoresbii, Pagophila alba, today, Proceedings of the 2nd International Con- all Lurus, Rhodostethiarosea, Rissa tridactyla. Crea- and all terns except Get- gress for Systematic and Evolutionary Biology. grusfurcatus, Xema sabint); (i.e., University of British Columbia, Vancouver. ochelidonnilotica Chlidoniasniger, Phaetusa sim- both SIBLEY, C. G., AND J. E. AHLQUIST. 1990. Phylogeny plex, Hydroprognecaspia, all , , and classification of birds. Yale Univ. Press, New Lurosterna inca, Gygis alba, and both ); Ryn- Haven, CT. chopsflavirostris; all Glareola except pratincola and nuchalis; Cursoriusrufw; all Burhinus except oedic- SIBLEY, C. G., J. E. AHLQUIST, AND B. L. MONROE, JR. and 1988. A classification of the living birds of the nemus,capensis, magnirostris;Esacus magniros- world based on DNA-DNA hybridization studies. tris; Phegornismitchellii; novaeseelandiae; Auk 105:409-423. Charadriusobscurus, montanus, coltaris, venustus, I$ icapillur, alticola, veredus,morinellus, sanctaehelenae, SNEATH, P.H.A., AND R. R. SOKAL. 1973. Numerical marginatus, cucullatus,semipalmatus, melodus, and taxonomy. W. H. Freeman, San Francisco. placidus; Vane&s melanopterus,lugubris, gregarius, SOBER, E. 1988. Reconstructing the past: parsimony, cinereus, resplendens,duvaucelii, macropterus, and evolution, and inference. Massachusetts Institute crassirostris;all Haematopus except ostralegus;Cla- of Technology Press, Cambridge, MA. dorhynchusleucocephalw, Recurvirostra americanu and SPRINGER, M., AND C. KRAJEWSKI. 1989. DNA hy- andina; and all auks (Endomychura hypoleuca, bridization in taxonomy: a critique from aalge, and Cepphusgrylle). Consequently, all of them first orinciples. @tart. Rev. Biol. 64:29 l-3 18. were marked as having missing data for characters 24- STRAUCH;J. G.; JR. 1976. The cladistic relationships 30 in my revision of Strauchs’ data matrix. of the Charadriiformes. Ph.D.diss. Univ. of Mich- igan, Ann Arbor, MI. APPENDIX 2: STRAUCH,J. G., JR. 1978. The phylogeny ofthe Char- adriiformes (Aves): a new estimate using the meth- REVISED STRAUCH DATA MATRIX od of character compatibility analysis. Trans. Zool. The data matrix employed in the present study is iden- Sot. Lond. 34~263-345. tical to that in Strauchs’ (1978) table 1, with the fol- SWOFFORD, D. L. 1991. PAUP: Phylogenetic Anal- lowing exceptions. vsis Using Parsimonv. Version 3.0s. Illinois Nat- ural History Survey, Champaign, IL. (1) The modifications described in the Revision of TIMMERMANN, G. 1957a. Studien zu einer verglei- Strauchs’ data matrix were implemented. chenden Parasitologie der Charadriiformes oder (2) Modifications to Strauchs’ matrix rendered some Regenpfeiferviigel. Teil I: Mallophaga. Parasito- of his terminal taxa identical. To avoid redundan- logische Schriftenreihe, Heft 8, Gustav Fischer, cy, identical terminals were subsumed under single Jena. taxon labels. These taxon labels, followed by the TIMME~MANN, G. 1957b. Stellung und Gliederung species they include, are: Jacanidae A (Actophilor- der Regenpfeifervijgel (Ordnung Charadriiformes) nis africana. Zrediparra gallinacea); Gallinago A nach Massgabe des Mallophagologischen Be- (G. megala, G. nigripennis);Gallinago B (G. ma- fundes, p. 159-172. In J. G. Baer [ed.], Premier crodactyla,G. media); NumeniusA (N. tahitiensis, symposium sur la specificit?. parasitaire des par- N. madagascariensis);Limosa A (L. limosa, L. asites de vertebres. Paul Attinger, Neuchatel. haemastica,L. lapponica);Phalaropus A (P. lob- 196 PHILIP C. CHU

atus, P. fulicarius); Actitis (A. ma&aria, A. hy- R. andina); CursoriusA (C. coromandelicus.C. poleucos);Calidris A (C. melanotos,C. acumina- temminckii); StercorariinaeA (Catharacta skua, ta); CalidrisB (C. pusilla, C. minuta); Stercorariuslongicaudus); Steminae A (Sterna hi- A (Charadriusmongolus, C. bicinctus,C. asiaticus, rundo. Anous minutus); Steminae B (Chlidonias Pluvialis dominica); Charadriidae B (Charadrius niger,Phaetusa simplex, Hydroprogne caspia, Ster- collaris, C. venustus,C. ruficapillus,C. alticola, C. na trudeaui);Larinae A (Larus scoresbii,PagophiIa veredus,Thinornis novaeseelandiae); Charadrius A eburnea,Larus Philadelphia, L. minutus, Rhodo- (C. marginatus,C, alexandrinusdealbatus); Hae- stethiarosea, Rissa tridactyla); and Iarinae B (Lar- matopusA (H. jinschi, H. moquini, H. frazari, H. us heermanni, L. delawarensis,L. argentatus,L. bachmani,H. ater);Recurvirostra A(R. americana, serranus,L. novaehollandiae,Creagrus furcatus).